Explainable machine learning analysis of medicine pricing and life expectancy across countries

Quantifying the statistical relationships between national medicine pricing structures and aggregate life expectancy is essential for shaping effective global health policy. This work utilizes an international panel dataset of 60 countries, incorporating 29 multidimensional health and socio-economic indicators, to evaluate the associations with medicine prices while controlling for confounding factors such as healthcare infrastructure, economic status, and environmental risk. Comparative analysis of seven machine learning algorithms identified Random Forest Regressor as the optimal model, achieving an R² of 0.77, RMSE of 2.90, and MAE of 2.02. Ablation experiments contrasting models with only structural controls, only price variables, and the combined specification confirmed that socio-economic factors explain most variance, while pharmaceutical pricing provides incremental predictive value by reducing error at the margin. Model explainability was enhanced through SHapley Additive Explanations (SHAP) and estimation of model-based effect estimates, which enabled category-specific interpretation across 13 therapeutic classes. Higher prices for corticosteroids, gastrointestinal, and respiratory drugs demonstrated negative associations with life expectancy in settings with limited affordability, whereas increased prices for endocrine, psychotropic, and antihistamine medications correlated with improved longevity in resource-rich health systems. Cross-country comparisons, particularly between India and Brazil, highlighted the influence of healthcare coverage and pricing policies in modulating these associations. These findings underscore the utility of explainable artificial intelligence in disentangling complex health-economic relationships and provide evidence to inform pharmaceutical pricing interventions for optimizing population health.